isGenAI.com

Glossary

EU AI Act Article 50: the transparency article

Article 50 of the EU AI Act (Regulation (EU) 2024/1689) sets transparency obligations for certain AI systems: providers must tell people they are interacting with AI and mark synthetic output in a machine-readable, detectable way, and deployers must disclose emotion recognition, deepfakes and certain AI-generated public-interest text, applying from 2 August 2026.

Definitions and isGenAI behaviour as of October 2026. Updated .

Read the Article 50 guidesFrequently asked questions
Code-drawn illustration for EU AI Act Article 50

Illustration · PNG file

  1. Signature + IHDR
  2. iTXt XML:com.adobe.xmprecord
  3. tEXt parameters
  4. tEXt prompt / workflow
  5. tEXt invokeai_metadata
  6. eXIf
  7. caBX C2PA manifest
  8. IDAT pixels
  9. IEND

AI source type declared

The file labels itself as AI-made.

IPTC DigitalSourceType such as trainedAlgorithmicMedia is read from XMP and shown with its exact value.

Check an image →
AI source type declared: where this record sits inside a PNG file and what the isGenAI report says about it. The picture is a code-drawn illustration.

The five paragraphs

Article 50(1) requires AI systems that interact with people to make that clear unless obvious. Article 50(2) requires providers of systems generating synthetic audio, image, video or text to mark outputs in a machine-readable format and make them detectable, using solutions that are effective, interoperable, robust and reliable as far as technically feasible. Article 50(3) covers deployer notices for emotion recognition and biometric categorisation. Article 50(4) covers deepfake disclosure and AI-generated text published on matters of public interest. Article 50(5) requires the information to be clear, distinguishable and timely.

The Commission has published guidelines and a code of practice on marking and labelling to support 50(2) and 50(4).

How isGenAI maps to Article 50

isGenAI’s Article 50 pages explain each obligation, the Article 50 scanner looks for disclosure signals on public products, the C2PA checker and provenance checker inspect real outputs for machine-readable marking, and the scorecards record observations with evidence hashes. The checker’s evidence classes map directly onto 50(2): a C2PA manifest, an IPTC source type and a saved record are all machine-readable marks, with different robustness.

What this site is not

These pages are practical explanations, not legal advice. The official text and the Commission’s guidance are linked; an organisation’s obligations depend on its role as provider or deployer and on its specific system.

How to use this term when checking a file

  1. Upload the original file to the AI image detector or the tool named on the page.
  2. Find the finding that uses this term under “How we checked” and read the source field it names.
  3. Weigh it by kind: a signed record, a declared field, a saved prompt and a matching copy are different strengths of evidence.
  4. Remember that absence of a record is not evidence of a human author.

Read the Article 50 guides

Frequently asked questions

When does Article 50 apply?

From 2 August 2026, per the AI Act’s application timetable.

What is machine-readable marking under 50(2)?

A provider-side mark that software can read, such as C2PA Content Credentials, IPTC metadata or a watermark, that must be effective, interoperable, robust and reliable as far as feasible.

Who must label a deepfake?

The deployer under Article 50(4), with narrow artistic and satirical exceptions.

Does isGenAI certify compliance?

No. It inspects files and public products and records observations with evidence. Legal assessment is separate.

Sources

Definitions follow the published standards linked above; the isGenAI behaviour described is the deployed checker as of October 2026.